The relevance of cut-stone to strategies for low-carbon buildings

3.1 Dynamic life cycle analysis (LCA)

To comply with recommendations by experts in the field (Levasseur et al. 2010, 2013), French regulations RE2020 LCA methodology, and likely upcoming European legislation (Ventura 2022), a dynamic LCA (dynLCA) approach was adopted.

The objective of this approach, which is increasingly sought after, is to lessen the critical lack of time dependence in traditional static LCAs by allowing consideration of the timing of emissions during the life cycle of a building. With a fixed time horizon (TH) of analysis (100 years for the most used GWP₁₀₀ unit), static LCAs (i.e. the sum of GWP₁₀₀ values of all emissions listed in the life cycle inventory of a product) effectively present inconsistencies between the chosen TH and the period covered by the results. The GWP₁₀₀ of emissions occurring 50 years after construction (e.g. end-of-life emissions) will, for instance, account for years 50–150 after construction. To remain consistent with chosen time boundaries, yearly GHG emissions (kg_GHG) reported in the life cycle inventory of a product are weighted within the dynLCA method via instantaneous dynamic characterisation factors (DCFs) derived for any GHG and emission timing relative to the TH of the analysis (Levasseur et al. 2010, 2013). However, typical EPDs lack differentiated life cycle emission inventory data, providing GWP₁₀₀ values (kg CO₂e) for each life cycle stage instead of yearly absolute emissions (kg_GHG), which precludes the use of standard dynLCA methodologies. To address this problem without making current EPD databases obsoletes, the RE2020 LCA methodology derives dynamic weighting coefficients (DWCs) from the residual atmospheric load following a pulse emission of 1 kg CO₂ given by the Bern carbon cycle model and directly applies them to GWP₁₀₀ values listed in life cycle inventories of EPDs (reasonably assuming CO₂ to be the main GHG constituent of embodied emissions for construction materials). As such, the total GWP of a product is weighted relative to the timing of each emission listed in the inventory.

As the RE2020 operates within statutory 50-year building life spans (BLSs) and a TH of 100 years, the decree only provides DWCs for GWP₁₀₀ factors and emissions taking place within the boundaries of LCI years 1–50. To allow sensitivity analyses both in terms of temporal boundaries and BLS, DWCs are derived for all analysis THs and LCI years (i), repeating equations (1) to (5) for a 1 kg CO₂ pulse emission incrementally applied to all inventory annual entries:

where DCF(t) is the dynamic characterisation factor used to assess CO₂ absorption that occurred t years before (W/year/m²); $a_{{\text{CO}}_2}$ is the instantaneous radiative forcing per unit mass of CO₂ increase in the atmosphere (W/m²/kg); $C{(t)}_{{\text{CO}}_2}$ is the atmospheric load of CO₂ t years after the emission (kg); $g_{{\text{CO}}_2}(i)$ is the inventory entry for year i; GWI_inst(t) is the instantaneous global warming impact occurring at a given time t; GWI_cum(t) is the cumulative global warming impact; and GWP_TH is the global warming potential at a given TH (i.e. traditional LCA unit; kg CO₂e) obtained by dividing the cumulative impact on global warming for this TH by the cumulative radiative forcing of a 1 kg CO₂ pulse emission occurring at time zero. Equations (1) to (3) were computed using the DynCO₂ Excel-sheet tool (CIRAIG 2013).

For clarity, Figure 1 illustrates the outcomes of the methodology for a 100-year analysis TH, with Figure 1(a) showing the cumulative radiative forcing of a pulse emission of 1 kg CO₂ in LCI year 1 according to the Bern carbon-cycle model, and Figure 1(b) showing the correlation between the timing of the emission and associated DWCs. The GWP₁₀₀ of an emission happening in LCI year 50 should, for instance, be weighted by a factor of 0.578 (i.e. 57.80%) in the LCA of a product given in a GWP₁₀₀ format, while emissions happening in LCI year 1 or 100 would respectively be fully considered (i.e. 100%) and fully discarded (i.e. 0%).

Figure 1

The LCI data for all subsequently considered materials were thus weighted with DWCs relative to the TH considered.

3.2 Cut-stones

LCI data for cut-stones are derived from all EPDs available to date on the French market and cradle-to-grave data compiled by contributors of the study (Pestre 2021) over the 2017–18 period as part of PhD field research, verified according to ISO 14025 (ISO 2020) and ISO 21930 (ISO 2023) (following NF EN 15804 recommendations) and examined by reviewers (Pestre 2021). All EPDs and datasets were further characterised by their density (kg/m³), thickness (m) and surface density (kg/m²). All datasets used in this section were configured in compliance with the NF EN 15804+A1 standard (note that industry feedback currently shows cut-stone GWP variations of less than 2% between NF EN 15804+A1 and +A2 standards; CEN 2019). Selected stages were revised to provide representative averages of the French stone quarrying and construction sector—which is analogous to certified EPD configurators for alternative materials later detailed (BETie 2022b; DE-Bois 2022a)—according to the following harmonising assumptions:

• A2 (product stages—transport to manufacturing site): median weighted value adopted for all datasets (function of surface density).

• A4 (construction process stages—transport to the building site): geological Geographic Information System (GIS) data were extracted from the Geological and Mining Research Bureau (BRGM) (2022) data platform to estimate national area coverage as a function of distance from potential quarries (cf. Figure 2). Limestone and sandstone quarries were considered (i.e. those most commonly employed in the construction sector), both active and closed (216 and 68 quarries, respectively), as both are assumed representative of the extraction potential of a given region. A conservative national average distance of 60 km was adopted, corresponding to an 85.3% national area coverage, comprising 92.3% of cities that have over 100,000 inhabitants. Truck loading rates were assumed to be 90%, with 30% empty return rates, corresponding to the latest observations showing heavy masonry transport to quickly reach maximum authorised loads (MEDDE & ADEME 2022), and aligning with the latest validated heavy masonry EPDs (CTMNC 2022) adapted from Ecoinvent allocations (MTECT 2021a; Wernet et al. 2021).

• A5 (construction process stages—installation into the building): inclusion of mortar, binding agents and secondary components (if any). Median weighted value adopted for all datasets (function of surface density), differentiating thin (< 20 cm) from thicker (> 20 cm) assemblies, as thin products are more likely to be installed with metal fasteners or more adhesive mortars than heavy masonry.

• B2 (use stages—maintenance): median weighted value (a function of surface density) adopted for all datasets (air-gumming every 50 years with an electric compressor may be required for aesthetic purposes only).

• C2 (end-of-life stages—transport to waste processing): average national transport distances of 60 km assumed (Figure 2a, b), with a partial raw material return to quarries for rehabilitation (see article R.515-3, Environmental Code 2023) and a partial return to the manufacturing site for size adjustments before reuse.

• C4 (end-of-life stages—disposal): 7.4% of landfill (i.e. industry average; MEEM 2017), and the rest considered without impact except for transport allocated to C2.

Figure 2

LCIs compiled for cut-stones (outlined in Appendix 2.1 in the supplemental data online) were then used to derive a dynamic model (see equation 1) for the volumetric carbon impact of cut-stone, providing GWP magnitudes for any TH (GWP_TH, in kg CO₂e/kg) as a function of cut-stone-product thickness (n, in m):

Subsequent works replicating the methodology should update the dataset with EPDs released after the completion of this study.

3.3 HWPs

LCIs for HWPs were derived from EPD configurable data certified by the MTE as representative of CLT panels manufactured in France in conformity with the NF EN 16351 standard (CEN 2021; DE-Bois 2022b; INIES 2022b), with industry-average contents of 84% and 16% of the mass, respectively, sourced from French and Scandinavian softwood forests. Four distinct methodologies were subsequently used to dynamically allocate the inventory data.

Following the RE2020 LCA methodology (RE2020 2021), the first scenario (hereafter referred to as CLT1) considers the sequestration of carbon to take place during the growth of the biomass (i.e. before the tree cut) and is allocated as a negative instantaneous pulse emission in LCI year 0.

By fully accounting for the year 0 sequestration and only partially accounting for year 50 end-of-life emissions (i.e. incineration, returning sequestered carbon to the atmosphere), DWCs specified by the RE2020 LCA methodology (also adapted from Levasseur et al. 2010, 2013) effectively turn biogenic products into carbon-negative commodities. Initially prescribed by the government to swiftly transition away from carbon-intensive cementitious materials by increasing demand for bio-based products (MTE 2020; MTES 2020; RE2020 2021), this sequestration allocation method nevertheless overlooks unsettled concerns by:

• clouding the nuanced short-term benefits of transferring carbon from one pool (forest) to another (building stock) (Arehart et al. 2021; Göswein et al. 2022; Hawkins et al. 2021)

• assigning a lower priority to complex risk assessments with reported trade-offs (notably in terms of biodiversity loss) from increasing demand for HWPs in the European market (Ceccherini et al. 2020; EFSOS 2011; Luyssaert et al. 2018)

• decoupling drivers for demand (i.e. RE2020 incentives) from projected national harvesting capacity (halving of the sequestration capacity of French forests between 2010 and 2022 due to the expansion of bio-aggressors and increased fire, hydric and climate stress (CITEPA 2022)), and decoupling drivers for imports from strong governance and careful planning required to ensure sustainable harvesting rates, i.e. longer rotation periods, and afforestation (EFSOS 2011; Mishra et al. 2022) within the framework of an increasing vulnerability of European forestry ecosystems to climate change (Au et al. 2022; Lindner et al. 2010)

• preventing differentiation of slow- versus fast-rotation biomass-based products and their respective climate disadvantages and benefits (Cherubini et al. 2011; Göswein et al. 2021; Guest et al. 2013; Levasseur et al. 2013; Pittau et al. 2018)

• neglecting reported uncertainties about sequestration dynamics at the individual tree scale (Calders et al. 2022; Mikoláš et al. 2021), showing rates of above-ground carbon accumulation increasing continuously with tree size (Stephenson et al. 2014), demonstrating the disproportionately important role of older trees in forest mass growth, and pointing to the overlooked effect of the avoided sequestration potential of managed forests prevented from attaining older ages (Franklin et al. 2002; Grantham et al. 2020; Moomaw et al. 2019).

To address some of these shortcomings and foster precaution, an alternate methodology was developed by Levasseur et al. (2013) for the case of sustainable forest management practices (as is assumed with CLT products considered here; DE-Bois 2022a; NF EN 2014), advocating for the dynamic (i.e. diffuse) allocation of sequestration intakes after the harvest during the regrowth of the tree (i.e. stemming from the assumption that a harvested tree entails an imminently replanted one). This allocation methodology, which was applied in subsequent studies (Göswein et al. 2021; Hawkins et al. 2021; Levasseur et al. 2013; Peñaloza et al. 2016; Pittau et al. 2018), is thus considered for a second scenario (referred to as CLT2). Detailed accounts of the considered rotation period, carbon fraction of dry biomass, and growth rate and allometric equations for French Norway spruce (Picea abies; 96% of total CLT mass in used CLT EPD configurator) are shown in Appendix 1.2 in the supplemental data online.

Avoided sequestration of the cut tree is systematically discarded from the GWP of HWPs, either stemming from the assumption that sequestration rates of the replanted tree quickly surpass that of the harvested one, or that trees have been planted with the incentive of using them as building wood products. In the context of climate urgency and growing pressures on forest resources, avoided sequestration should ultimately be considered to provide a comprehensive perspective on the short TH climate impact of harvesting activities and allow fair inter-material comparative assessments. The second allocation scenario (CLT2) is thus further supplemented with avoided sequestration dynamics, resulting in two additional scenarios (hereafter referred to as CLT3 and CLT4) where avoided sequestration is accounted for until the yearly sequestration rates of the replanted tree surpass that of the cut tree had the latter not been cut (i.e. until the replanted tree provides net climate benefits). As uncertainties about Norway Spruce growth rate dynamics remain, two distinct models were used.

The first high-probability scenario (CLT3) considers a standard growth rate model (Figure 3b) (Albers et al. 2019; Vallet et al. 2006) with sequestration rates decreasing with tree size. A second lower-probability scenario, CLT4 (i.e. a model representative of growth in the absence of competition) (Figure 3a) considers a metabolic scaling model (Stephenson et al. 2014) where growth rates increase continuously with tree size. Both models are fitted to account for 443 kg of dry mass (corresponding to the biogenic material mass in 1 m³ of CLT as detailed in underlying EPD data) at LCI year 82 (i.e. the assumed forest rotation period; Eriksson et al. 2007; Seely et al. 2002). As the metabolic scaling model does not account for a maximum tree dry mass, the latter is capped to a final dry mass equivalent to that of the standard model (700 kg, reached in LCI year 104 and 200 of the metabolic and standard and models, respectively).

Figure 3

The dynamic life cycle inventories compiled for HWPs are outlined in Appendix 2.2 in the supplemental data online.

3.4 Concretes

LCIs for concretes were derived from EPD configurable data certified by the Ministry of Ecological Transition as compliant with NF EN-206/CN standard (CEN 2014b; BETie 2022a, 2022b; INIES 2022b). Concrete classifications and exposure classes are described according to the NF EN 197-1 (CEN 2012) and NF EN 206/CN standards. Two specific concrete types were considered as representative of conventional (100% clinker content—the reference scenario) and low-carbon (35–64% clinker substitution by GGBFS) load-bearing (20–25 MPa) reinforced concrete elements (i.e. C20 XC1 CEM I and C20 XC1 CEMIII/A, respectively). A total of 50 kg/m³ of reinforcing bars are assumed in each case, which follows industry averages (BETie 2022a, 2022b). A conservative value of 30 km, the national average (Guiraud 2018), was considered for the transport distance from the plant to the building construction site (life cycle stage A4).

As a co-product of the iron and steel industry, the allocation method of GGBFS emissions falls under the specifications of the NF EN 15804/+A1 standard for derivatives, subject to interpretation and enabling either mass, economic, physical or energetic attribution. Two alternative allocation methodologies were considered, resulting in two distinct low-carbon concrete scenarios.

Complying with the RE2020 LCA methodology and a recent arbitrage by the Directorate of Housing, Urban Planning and Landscapes (DHUP) (2022), a first low-carbon concrete scenario (hereafter referred to as CEM III 1) assumes the currently statutory economic allocation of 1.40%, resulting in a GGBFS emission factor of 0.083 kg CO₂e/kg_GGBFS (DHUP 2022) (detailed accounts of average European market prices, mass correlation factors and emission factors are shown in Appendix 1.3 in the supplemental data online).

As economic allocations index the environmental impact of processes to their financial worth at the macroeconomic level with little correspondence to their ecological reality, international LCA standards (ISO 14044; ISO 2022) warn about the shortcomings of such procedures and advise, when feasible, to systematically adopt natural-science-based approaches (privileging direct or mass allocation solutions, in that order). Using empirical data from 30 blast furnaces of the ArcelorMittal group, a direct allocation for GGBFS was derived by Buttiens et al. (2016) using a differential approach, and evaluated at 0.55 kg CO₂e/kg_GGBFS (standard deviation (SD) = 0.018 kg CO₂e/kg_GGBFS, i.e. ±3%). This benchmark value, close to a mass allocation (i.e. 0.35 kg CO₂e/kg_GGBFS), is thus used for a second low-carbon concrete scenario (CEM III 2).

Adopted carbonation kinetics are detailed in Appendix 1.3 in the supplemental data online, and dynamic LCIs compiled for all considered scenarios are listed in Appendix 2.2 online.

3.5 Functional unit

The functional unit used for the comparative analysis (Figure 4) consists of a 1 m² load-bearing wall (whether interior or peripheral). As typically implemented characteristics depend on product specificities offered by the industry (i.e. required properties rounded up to the closest performing available product) more than absolute required functionalities (i.e. exact required properties targeted), this study adopted a practice-based (as opposed to performance-based) functional unit.

Figure 4

Practice-based characteristics (here typically implemented thicknesses) were derived in a two-step process. First, archetype-based data were derived from the statistical analysis of 1.2 million housing units constructed in France over the period 2013–22, with data retrieved from the Sitadel2 database (Sitadel2 2022). Average collective housing building heights above ground level are shown to be 3.5 ± 1.76 stories (SD). Collective housing projects currently represent 74% of new developments owing to strong government commitment to halt urban sprawling (Légifrance 2021) and were thus assumed representative of dominant practices. Average building heights were then cross-referenced with correlated industry-average load-bearing wall thicknesses, amounting to 13 cm for CLT (BBG 2020; Egoin 2018; KLH 2013), 20 cm for concrete (2–3 m-high wall average thickness; CIMbéton 2013) and 24 cm for cut-stone (Pestre 2021)—all exhibiting compressive strengths greater than the design loads anticipated for the considered building type.¹

Foundation design and the correlated embodied carbon are (1) relatively independent from variation in load-bearing wall surface density (e.g. estimated 7.5% variation range of foundation concrete volume for all considered assemblies, assuming French archetypal collective housing building characteristics, and Eurocode 7 section 6 requirements for strip footing design); (2) highly dependent on soil properties (Eurocode 7); and (3) effectively discarded from LCA system boundaries within the new French building code RE2020 LCA methodology (i.e. capping of below-ground carbon impact). As such, the functional unit discards the relative impacts of the various wall systems on the foundations.

Unlike interior partitions, peripheral walls are submitted to insulation requirements. In the latter case, the difference in thermal resistance between the different materials and considered wall thicknesses would be relatively insignificant with respect to that required to comply with the RE2020. Average R-values for 13 cm CLT and 24 cm stone (1.08 and 0.27 m².K/W, respectively, against the 6.5 indicated to satisfy RE2020 requirements) would, for instance, result in a change in insulation thickness of 2.7 cm (18.3 and 21.0 cm required for CLT and stone, respectively²), for an average insulation product thicknesses of 9.1 cm on the French market (INIES 2022a). As such, the impact of differing thermal resistance on the functional unit is also discarded.

As all considered assemblies comply with average acoustic insulation requirements of 30dB (Légifrance 1999)—i.e. mass law approximations of 36, 58 and 56 dB insulation for CLT, concrete and cut-stone considered thicknesses, respectively (Li & Ren 2011)—the impact of sound performance is discarded from the functional unit.

As implemented wall thicknesses may vary according to project-specific construction methods and technical requirements, the results presented hereafter (i.e. Figure 8) show the embodied impact as a function of wall thickness, with a marker to distinguish the specific thickness considered in the functional unit, making it possible to compare different scenarios.

The 1 m² unit was then scaled up to the level of French building parc dynamics using the approach outlined in the subsequent section.

3.6 Building stock model

The state-of-the-art French dynamic building stock model developed by Blanchet et al. (2021) is used to derive projected yearly construction rates (i.e. trend scenario) for the 2025–50 period. Statistical archetype-based data (European TABULA project; Loga et al. 2016) are used to obtain average wall geometry factors for collective housing projects, assumed to be representing an average of 67.5% of all future constructions over the period (ADEME 2021c). At the parc level, the analysis subsequently centres on examining the comparative benefits of implementing the various material alternatives on 30% of new collective housing projects during the period (which would generate a peak demand for cut-stone of 870,988 m³/year, expected to occur in 2035). Dynamic building stock variables and material LCIs over the considered period are provided in Appendix 2.3 in the supplemental data online.

4.1 Embodied carbon of cut-stone

GWP values for the volumetric carbon impact of cut-stone were derived on a per kg of stone basis, as a function of material thickness. A power function model (equation 7) (plotted in Figure 5 for TH = 100 years and BLS = 50 years) is found to best fit the distribution of the data. The following observations can be made:

• The model distinctly displays decreasing trends (i.e. lower impact per kg of stone for higher thicknesses). Similar to previous findings on isolated products (Ioannidou et al. 2014), this observation coincides with the fact that sawing and surface treatment emissions are correlated with the specific surface of the product, regardless of its volume. Thin assemblies are also more likely to be composed of higher density stones, necessitating longer sawing times, and to require secondary products for installation (e.g. metal fasteners, accounted for in LCA stage A5), as well as more adhesive mortars, which are emission intensive.

• Higher amplitudes of variability, and thus lower predictability, are found for thinner products (n < 0.2 m) due to the unpredictable nature of the required add-ons and the increasing variability in surface treatment procedures (e.g. flaming or polishing more likely to be undertaken for cladding than for heavy masonry walls, often left with cut marks).

Figure 5

Consequently, the model is distinctively evaluated for thicknesses above and below 0.2 m. The separate assessments show varying root mean square errors (RMSEs) for the two ranges: while model performances are lower for thinner assemblies (RMSE = 0.20 kg CO₂e/kg), the reliability noticeably increases (15-fold) for assemblies > 0.2 m (RMSE = 0.013 kg CO₂e/kg, equivalent to 5.77 kg CO₂e/m² of 0.24 m-thick wall for average densities of the French quarrying sector; see section 4.2), above which the model is reasonably deemed robust enough to accurately position heavy masonry walls among other material alternatives in the subsequently undertaken comparative LCA.

Accordingly, for any TH, the model assumed in the subsequent analyses takes the following form:

where n is the thickness of the product (m); and the scaling factor (δ) and exponent (ɛ) are derived for each TH and BLS (Table 1).

The distribution of emissions among the various life cycle stages for thicker stone material assemblies is shown in Figure 6 for the collected data, with product and construction stages (A1–A5) accounting for more than three-quarters of the total cumulative impact.

Figure 6

A transport distance (life cycle stage A4) sensitivity analysis is performed. Assuming average French limestone densities (Pestre 2021) and typically implemented thicknesses of 24 cm, the total embodied impact of a masonry wall (1 m²) is shown to double for truck-transit distances increasing from 60 to 579 km, in which case the A4 stage would go from representing 12% to 56% of the total impact. A similar increase in transport distances when considering electric rail transport would only increase the total impact by 5.3%, suggesting that national coverage from selected quarries with high extraction rates would be an appropriate strategy for national low-carbon transition industrial roadmaps. Although overseas importations (via ship transit) are mostly limited to thin products (SNROC et al. 2014), cross-border competition has been shown to arise within the EU for thicker assemblies, with truck transit sometimes exceeding 1500 km (INIES 2020). While economic flow analyses have shown such importations to be unsuitable strategies in social terms (Ioannidou et al. 2018), the results also indicate their unsuitability from an environmental perspective.

4.2 Comparative LCA

Using a French average limestone density of 1851 kg/m³ (Pestre 2021), the volumetric carbon-impact model for cut-stone was used to comparatively assess the performances of load-bearing stone masonry walls with regard to low-carbon alternative materials in the context of the French construction sector.

Following RE2020 requirements (RE2020 2021), results are first shown (Figure 7a and Table 2) for a 50-year BLS and a 100-year analysis TH (i.e. GWP₁₀₀). The GWP scale is shown for functional units of a 1 m² wall. Uncertainty ranges are indicated in red, light grey and grey for stone, HWPs and low-carbon concrete, respectively. Higher probability scenarios are indicated with increased line weight, corresponding to the following:

• median estimation range of the model in the case of stone (referred to as STONE 1)

• CEM III with direct attribution of GGBFS emissions in the case of low-carbon concrete (referred to as CEM III 2)

• sequestration accounted for during regrowth and avoided sequestration accounted for using a decreasing growth rate model (Figure 3b) in the case of HWPs, referred to as CLT3—the metabolic rate model CLT 4 having been developed for tree growth in the absence of competition (Stephenson et al. 2014), representative of timber sourced from natural forests newly turned into a managed forests, as opposed to a managed forest established through afforestation.

Figure 7

Due to their volumetric impact decreasing with increasing thickness, stones are shown to present the lowest absolute increase rate in GWP₁₀₀ with increasing wall thickness (30.9% GWP₁₀₀ increase for a doubling of material volume compared with 212.0% and 200.0% for concretes and HWPs, respectively).

For high-probability material emission profiles, and considering typically implemented material thicknesses, stone assemblies will emit respectively 3.06, 2.73 and 1.43 times less than CEM I, CEM III and CLT walls. Considering the higher range of uncertainty for all materials, and typically implemented thicknesses, stone assemblies could emit up to 3.62 times less than HWPs. A transport distance sensitivity analysis is shown in Figure 8, considering average truck, ship and train transit (Ecoinvent 3.7.1). Assuming industry average transport distances between manufacturing and construction sites (LCA stage A4) for concrete and CLT (i.e. 30 and 287 km, respectively) (DE-Bois 2022a; Guiraud 2018), and GWP₁₀₀ values, truck-transit transport distances for cut-stone would have to be increased from the considered average of 60 km to 259, 1334 and 906 for its embodied impact to surpass that of the CLT3 (high-probability), CLT4 (low-probability) and CEM III 2 (high-probability) scenarios, respectively, suggesting that lorry-carried cut-stone imports, e.g. typically > 1500 km for neighbouring countries cut-stone products (INIES 2020), would quickly reach impact levels higher than locally sourced CLT and concrete. On the other hand, assuming electrically powered train transit, transport distances of 1721, 8914 and 6012 km would be required to reach embodied impacts higher than locally sourced CLT3, CLT4 and CEM III 2, respectively, suggesting that under clean nationwide transportation scenarios, cut-stone assemblies could consistently exhibit better performance.

Figure 8

It should be noted that transport distances to the manufacturing site (LCA stage A2) may also disproportionately vary between the considered materials, with CLT manufacturers increasingly relying on long-distance imports from Northern Europe’s boreal forests, against the common juxtaposition of extraction and manufacturing sites for cut-stone (ADEME 2016). With average increases of 7.54 kgCO₂e (GWP₁₀₀)/m² of 13 cm-thick CLT panel for 100% Scandinavian softwood contents (DE-Bois 2022a), truck-transit transport distances for 24 cm-thick cut-stone assemblies would have to be increased to 465 km to surpass that of 13 cm-thick CLT exclusively sourced from Northern boreal forests, in the CLT3 scenario, and to 1551 km in the CLT 4 scenario, suggesting that national truck coverage from a single quarry might provide a viable alternative to imported softwood-based products, from a carbon standpoint.

A sensitivity analysis was performed for a longer BLS (100 years) to further extend the comparative analysis. An increased TH was required for the analysis (200 years) to dynamically account for end-of-life emissions in LCI year 100. GWP₂₀₀ values are given in Figure 7(b) and Table 2. The results show that variation in the BLS has a minimal impact on geo-sourced and cementitious materials. Increased maintenance requirements result in a 1.01% average increase in the embodied impact of stone assemblies, while increased carbonatation (an exponentially decaying process) results in a 1.05% average decrease in the embodied impact of concrete assemblies. Conversely, the BLS is shown to greatly influence the comparative performance of HWPs. By delaying the end-of-life emissions of the CLT assembly and thus storing carbon in the building for longer periods, the relative significance of the diffuse effect of sequestration during the regrowth of the tree increases proportionately. While stone will emit on average 1.43 times less than CLT (high-probability scenario, CLT3) for typically implemented thicknesses and a BLS of 50 years (Figure 7a), the trend reverses for a life span of 100 years (Figure 7b), with stone emitting on average 3.20 times more than CLT. A dynamic BLS sensitivity analysis provided in Appendix 1.4 in the supplemental data online shows that the CLT3 scenario becomes less impactful than stone for any BLS above 85 years, suggesting that the BLS must be greater than the forest rotation period for the CLT to have a lower impact than stone.

However, this assertion is only valid if distant THs (100 years for Figure 7a and 200 years for Figure 7b) are selected for analysis. Climate urgency requires the temporal dynamics of material emission profiles to be considered for an astute understanding of their immediate impact.

To that end, the temporal dynamics of the two analyses (50- and 100-year BLSs) are shown in Figure 9 for typically implemented thicknesses, where the GWPs of the various assemblies are dynamically characterised for incremental THs. For further clarity, the correlation between Figures 7 and 9 is presented in Figure 10. Consistent with the scientific consensus (Hawkins et al. 2021), the results show that HWPs are not climate neutral in the short term, affecting their immediate comparative performance. The 100-year BLS analysis (Figure 9b) shows that the peak impact of CLT (high-probability scenario CLT3) occurs 15 years after construction, where the GWP₁₅ is 1.69 times that of stone. CLT (high-probability scenario CLT3) is shown to reach an impact lower than that of stone in year 38. The higher range of uncertainty for CLT (CLT4), on the other hand, is shown to never reach GWP levels lower than that of stone, with a short-term peak difference in year 37 showing a 3.04 times poorer performance compared with stone. Unlike HWPs, and due to the majority (77.4%) of the life cycle emissions being concentrated on the product and construction stage (LCA stage A) (Figure 6), the emission profile of cut-stone wall assemblies is shown to remain constant through time, with average long-term impact (TH = 100) shown to surpass short-term impact (TH = 10) by only 8.3%.

Figure 9

Figure 10

The comparative impact of transport distances (A4) in the early life cycle stages (i.e. with a short-term impact) should also be analysed in light of the temporal dynamics of the emission profiles. The peak impact exhibited by the CLT3 scenario in year 15 (Figure 9a) corresponds to the maximum divergence from the median-stone scenario. Considering GWP₁₅ values, cut-stone transport distances would need to increase from 60 to 351 and 556 km to exceed the impact of CLT sourced from French and Northern Europe’s boreal forests, respectively (assuming the CLT3 scenario for both cases). This suggests that the relevance of nationwide truck coverage from a single quarry may be debatable, at least from an immediate impact perspective.

At the level of the parc, the emissions associated with the implementation of the considered assemblies on 30% of new collective housing projects over the 2024–50 period are dynamically allocated for BLSs of 50 years (Figure 11a series), and 100 years (Figure 11b series). The resulting yearly instantaneous and cumulative global warming impacts (GWI_inst and GWI_cum, respectively) are plotted in the first and second rows of Figure 11 (see equations 2 and 3), while the relative impact (dynamic GWP) is shown in the third row. Previously observed trends are accentuated.

Figure 11

Considering high-probability emission profiles and statutory 50-year BLSs, the substitution of 0.2 m-thick CEMI with 0.24 m-thick cut-stone on the considered fraction of the parc would amount to a 2.77 Mt CO₂e decrease in the GWP₁₀₀ impact of new constructions (i.e. dynamically weighted relative to 2025), against 0.43 for CEM III and 1.18 for CLT, with the stone scenario presenting higher comparative benefits both in the short and long terms. Considering the higher range of uncertainty for all materials, the implementation of stone over CLT could yield savings of up to 5.43 Mt CO₂e over the period.

The 100-year BLS analysis shows peak short-term differences between CLT (high-probability CLT3 scenario) and cut-stone occurring in 2058, where GWP₃₃ magnitudes are 2.33 times higher for CLT, amounting to a 1.09 Mt CO₂e difference. The relative impact of the stone scenario (STONE 1) is shown to become higher than that of the CLT3 scenario in 2080, suggesting short-term climate benefits independent from BLSs for the former.